In mid-2024, OpenAI introduced a multi-modal GPT-4-o which can generate text, images and audio. It is an improvement over the previous system such as DALL-E3. It handles more complex requests with great accuracy. The prompts are translated into art and the model is capable of mimicking distinct art styles with great fidelity.
Over the concluding week of March 2024 and early April 2024, the model mimics the signature animation style by studio Ghibli, Inc. — a Japanese studio based in Koganei, Tokyo. Users have been tinkering with the model to test its capabilities. They realized that the AI tool has the ability to recreate Studio Ghibli’s hand-drawn art style. The internet is flooded with Ghibli-style user selfies. Ordinary objects such as cell phones, pictures of people with their pets, the fall of twin towers in 2001 — all these appear on internet. Even Sam Altman’s profile picture was converted on OpenAI into a Ghibli version.
The OpenAI model experienced outrages due to massive surge in traffic with users trying to create Ghibli-esque images.
There has been a huge debate on creativity, consent and artistic integrity. Perhaps, OpenAI model has been unfairly trained upon the copyrighted material and devalues the creative product of artists. It is all done without permission.
Besides this, sensitive subjects are reduced to cozy art.
AI models trained on private photos without consent can be misused. These can reveal metadata, location details and others sensitive information.
Leave a Reply